Capacity and Sizing - Consolidation Guide¶
1. Overview¶
Purpose: This document consolidates ALL capacity planning information across Webex Contact Center components. Rather than duplicate existing detailed content, this serves as a master reference guide pointing to comprehensive capacity planning already documented, while adding NEW capacity considerations for AI/automation and storage.
Document Strategy¶
This document provides: - ✅ Quick reference to existing detailed capacity planning (with links) - ✅ NEW detailed content for AI/automation capacity (not covered elsewhere) - ✅ NEW detailed content for storage capacity planning - ✅ Consolidated view of all capacity requirements in one place
Use This Document To:¶
- Get a complete picture of ALL capacity requirements
- Find where detailed capacity planning is documented
- Understand AI/automation capacity considerations (new)
- Plan storage and retention capacity (new)
1. Overview¶
Purpose: Master reference for ALL capacity planning across Webex Contact Center. This document: - 📍 References existing detailed capacity planning (avoids duplication) - 📝 Provides NEW content for AI/automation, storage, and IVR capacity - 🎯 Consolidates all capacity requirements in one place
2. Capacity Planning Quick Reference¶
2.1 Voice/Telephony Capacity COVERED EXTENSIVELY¶
📍 See: cube-and-sbc-design.md - Section 4 (380 lines, most comprehensive)
What's Covered: - ✅ CUBE session capacity formulas - ✅ Encryption impact calculations (TLS+SRTP = ÷3 capacity) - ✅ Example: 1,000 agents → 6,084 sessions required - ✅ Hardware sizing: 2× ASR 1002-HX (7,000 sessions total) - ✅ Reusing existing CUBE decision matrix - ✅ Session sizing worksheet - ✅ PSTN trunk capacity
Quick Summary:
Formula: ((Agents × 2) + Queue Calls) × 3 (encryption buffer)
Example: ((1,000 × 2) + 70) × 3 = 6,210 sessions
Hardware: 2× Cisco ASR 1002-HX = 7,000 sessions
Headroom: 13% buffer for growth
2.2 Network Bandwidth Capacity COVERED EXTENSIVELY¶
📍 See: network-architecture.md - Section 11 (Network capacity planning)
What's Covered: - ✅ Internet circuit sizing (dual ISP, 500 Mbps each) - ✅ Voice bandwidth: 87 Mbps for 1,000 concurrent calls (G.711 codec) - ✅ Video bandwidth: 75 Mbps for 50 sessions - ✅ Agent desktop: 500 Mbps for 1,000 agents - ✅ QoS headroom and oversubscription - ✅ Home agent bandwidth requirements - ✅ Capacity growth planning table
Quick Summary:
Total Required for 1,000 Agents:
├─ Voice: 87 Mbps (1,000 calls @ G.711)
├─ Video: 75 Mbps (50 concurrent sessions)
├─ Desktop: 500 Mbps (1,000 agents @ 0.5 Mbps)
├─ Mgmt: 20 Mbps (monitoring, management)
└─ TOTAL: 682 Mbps → Provision: 1,000 Mbps (dual 500 Mbps ISPs)
2.3 Agent Licensing Capacity COVERED¶
📍 See: assumptions-and-dependencies.md - Section 4.2 (License requirements)
What's Covered:
- ✅ License requirements table (Premium, Standard, Supervisor)
- ✅ 1,000 Premium agents @ $120/user/month
- ✅ 50 Supervisors @ $90/user/month
- ✅ 1,000 Webex Calling @ $25/user/month
- ✅ Recording storage: 500 hours/month @ $0.10/hour
- ✅ CUBE session license: 2,000 sessions (one-time)
- ✅ Total annual cost: ~$1,850,000
Quick Summary:
Agent Capacity Planning:
├─ Current: 1,000 agents
├─ Design: 1,300 agents (30% growth buffer)
├─ Peak: 700 concurrent (70% occupancy)
└─ Licenses: 1,050 (buffer for training, attrition)
2.4 Agent Capacity by Location COVERED¶
📍 See: design-principles.md - Section 4.1 (Agent capacity)
What's Covered: - ✅ Current: 1,000 agents across 3 locations - ✅ Design capacity: 1,300 agents (30% buffer) - ✅ Location breakdown: - Hyderabad: 600 agents - Austin: 300 agents - London: 100 agents
3. IVR Capacity Planning (NEW CONTENT)¶
3.1 Traditional IVR Port Capacity¶
IVR Port Sizing Formula:
IVR Ports = (Call Arrival Rate × Avg IVR Duration) / 60
Example:
- Calls per hour: 1,200
- Avg IVR time: 90 seconds (1.5 minutes)
- Required ports: (1,200 × 1.5) / 60 = 30 IVR ports
- With 20% buffer: 36 ports
Webex Connect Pricing Model: - Pay-per-use (no fixed port licensing) - Concurrent session-based - Elastic scaling (cloud-native)
3.2 Speech Recognition (ASR) Capacity¶
Concurrent Speech Sessions:
For 1,000-agent deployment:
├─ Peak IVR calls: 300 concurrent (30% of agents)
├─ Speech-enabled calls: 60% use ASR = 180 sessions
├─ With buffer (20%): 216 concurrent ASR sessions
└─ Provider: Google Speech-to-Text (1,000 streams/region)
Google Speech-to-Text Quotas: | Quota Type | Default Limit | Recommended for 1,000 Agents | |------------|---------------|------------------------------| | Concurrent streaming requests | 100/region | Request increase to 250 | | Synchronous requests/min | 1,000 | Sufficient | | Total audio minutes/day | Unlimited | ~10,800 minutes/day |
Cost Estimate: - Standard model: $0.006/15 seconds = $0.024/minute - 180 concurrent × 2 min avg × 60 min/hour × 12 hours = 259,200 minutes/day - Daily cost: 259,200 × $0.024 = $6,221/day or $186K/month
3.3 IVR Database Query Capacity¶
Backend Integration Load:
For account lookup IVR flows:
├─ IVR calls/hour: 1,200
├─ DB queries per call: 2 (lookup + verification)
├─ Total queries/hour: 2,400
├─ Peak queries/second: 0.67 QPS
└─ Database capacity: Need <100ms response time
Recommendations: - Database connection pool: 50 connections - Query timeout: 5 seconds - Caching for frequent lookups (reduce DB load)
4. AI and Automation Capacity Planning (NEW CONTENT)¶
4.1 Virtual Agent (Chatbot) Capacity¶
Concurrent Bot Sessions:
For 1,000-agent contact center:
├─ Digital deflection target: 30% of calls
├─ Peak calls/hour: 1,500
├─ Deflected to bot: 450 calls/hour
├─ Avg session duration: 4 minutes
├─ Concurrent bot sessions: 450 × (4/60) = 30 sessions
├─ With growth buffer (50%): 45 concurrent sessions
└─ Dialogflow CX capacity: Sufficient (scales automatically)
Dialogflow CX Quotas: | Quota | Default | Required for 1,000 Agents | Action | |-------|---------|---------------------------|--------| | Requests/minute | 600/project | ~100/min (peak) | Sufficient | | Active sessions | 1,000 | 45 concurrent | Sufficient | | Projects | 50 | 1-3 (dev/test/prod) | Sufficient | | Maximum session length | 30 minutes | 10 min avg | Sufficient |
Cost Estimate (Dialogflow CX): - Text requests: $0.007/request - 450 sessions/hour × 10 requests/session = 4,500 requests/hour - Monthly: 4,500 × 12 hours × 22 days = 1.2M requests - Cost: 1.2M × $0.007 = $8,400/month
4.2 Natural Language Understanding (NLU) API Capacity¶
Intent Detection Load:
Voice + Chat + Email combined:
├─ Voice IVR (NLU): 180 concurrent sessions
├─ Chatbot: 45 concurrent sessions
├─ Email auto-classification: 50 emails/hour = 0.014 concurrent
├─ Total NLU requests/min: ~90/min (peak)
└─ Quota needed: 600/min (default sufficient)
Dialogflow CX Rate Limits: - Default: 600 requests/minute/project - Recommended: Request increase to 1,000/min for production - Burst handling: Cloud-native auto-scaling
4.3 Real-Time Call Transcription Capacity¶
Live Transcription for Agent Assist:
Selective transcription strategy:
├─ Total concurrent calls: 700 (70% of 1,000 agents)
├─ Transcribe high-value calls: 20% = 140 calls
├─ Provider: Google Speech-to-Text (streaming)
├─ Quota needed: 1,000 concurrent streams/region
└─ Status: Within quota limits
Cost Estimate (Real-Time Transcription): - Enhanced model: $0.009/15 seconds = $0.036/minute - 140 calls × 8 min AHT × 60 min/hour × 12 hours = 806,400 minutes/day - Daily cost: 806,400 × $0.036 = $29,030/day or $871K/month
💡 Recommendation: Start with 50 calls (10%) to control costs
4.4 Sentiment Analysis Capacity¶
Post-Call Sentiment Analysis:
Analyze all calls (batch processing):
├─ Daily calls: 10,000 calls
├─ Transcription required: 10,000 calls × 8 min = 80,000 min
├─ Sentiment analysis: 10,000 API calls
├─ Provider: Google Natural Language API
└─ Processing time: ~6 hours (batch overnight)
Google Natural Language API Quotas: | Quota | Default | Required | Status | |-------|---------|----------|--------| | Requests/minute | 600 | ~100/min (batch) | Sufficient | | Requests/day | 800,000 | 10,000/day | Sufficient |
Cost Estimate: - Sentiment analysis: $1.00/1,000 text records - 10,000 calls/day = 10,000 records - Daily cost: 10,000 × $1.00/1,000 = $10/day or $300/month
4.5 Agent Assist - Knowledge Base Capacity¶
Real-Time Knowledge Searches:
Agent assist during calls:
├─ Concurrent calls: 700 (70% of agents)
├─ Knowledge searches/call: 3 searches
├─ Calls with searches: 30% = 210 calls
├─ Total searches/hour: 210 × 3 = 630 searches/hour
├─ Searches/minute: 10.5/min
└─ Search latency required: <2 seconds
Webex Agent Answers Capacity: - Cloud-based (Cisco-managed) - Scales automatically - No explicit capacity limits - Charged per agent license
Knowledge Base Size: - Articles: 5,000-10,000 articles - Index size: ~500 MB - Search index: Elasticsearch (cloud-hosted)
4.6 Post-Call Analytics - Batch Processing¶
Speech Analytics Platform Capacity:
For CallMiner/NICE/Verint:
├─ Daily recordings: 10,000 calls
├─ Avg call length: 8 minutes
├─ Total audio: 80,000 minutes/day = 1,333 hours
├─ Processing speed: Real-time × 10 (10 min audio = 1 min processing)
├─ Required capacity: 133 hours processing/day = 5.5 hours actual
└─ Batch window: Overnight (12 hours) - Sufficient
Storage for Analytics: - Audio recordings: 1.6 TB/day (64 kbps × 10,000 calls × 8 min) - Transcripts: 200 GB/day (text) - Metadata: 50 GB/day - Total: 1.85 TB/day or 56 TB/month
4.7 AI/ML Model Training Capacity¶
Training Data Volume:
For predictive routing model:
├─ Training dataset: 100,000 historical calls
├─ Features per call: 50 attributes (duration, sentiment, outcome, etc.)
├─ Dataset size: ~5 GB
├─ Retraining frequency: Weekly
├─ Training duration: 2-4 hours
└─ Compute: Cloud ML (Google AI Platform or AWS SageMaker)
Cost Estimate (Model Training): - Google AI Platform: $2.50/hour (training) - Weekly training: 4 hours × $2.50 = $10/week - Monthly cost: $40/month
4.8 API Rate Limits Summary (Critical!)¶
Third-Party AI Service Limits:
| Service | Default Quota | Required for 1,000 Agents | Action Needed |
|---|---|---|---|
| Dialogflow CX | 600 req/min | 100 req/min | ✅ Sufficient |
| Google Speech-to-Text | 100 concurrent streams/region | 250 streams | 🔴 Request increase |
| Google Natural Language | 600 req/min | 100 req/min | ✅ Sufficient |
| OpenAI GPT (if used) | 3,500 tokens/min | Varies | 🟡 Monitor usage |
| Webex Connect APIs | 1,000 req/min | 200 req/min | ✅ Sufficient |
💡 Key Action Items: 1. Request Google Speech-to-Text quota increase to 250-300 concurrent streams 2. Monitor Dialogflow CX usage and request increase if approaching 600 req/min 3. Implement rate limiting and retry logic in all AI integrations 4. Set up quota monitoring alerts (at 70% threshold)
4.9 AI Capacity Cost Summary¶
Monthly AI/Automation Costs (1,000-agent deployment):
| Component | Monthly Cost | Annual Cost |
|---|---|---|
| IVR Speech Recognition (ASR) | $186,000 | $2,232,000 |
| Real-Time Transcription (10% calls) | $87,100 | $1,045,200 |
| Dialogflow CX (Chatbot) | $8,400 | $100,800 |
| Sentiment Analysis (post-call) | $300 | $3,600 |
| Model Training | $40 | $480 |
| TOTAL | $281,840 | $3,382,080 |
💡 Cost Optimization: - Start with 10-20% call transcription (not 100%) - Use batch processing instead of real-time where possible - Implement caching for frequent queries - Use standard ASR models (vs enhanced) where accuracy permits
5. Storage Capacity Planning (NEW CONTENT)¶
5.1 Call Recording Storage¶
Storage Calculation:
For 1,000-agent deployment:
├─ Daily calls: 10,000 calls
├─ Avg call duration: 8 minutes
├─ Recording format: 64 kbps compressed audio
├─ Daily storage: 10,000 × 8 min × 64 kbps = 1.6 TB/day
├─ Monthly storage: 1.6 TB × 22 days = 35.2 TB/month
├─ Retention policy: 90 days (compliance)
└─ Total required: 35.2 TB × 3 = 105.6 TB
Storage Tiers: | Tier | Retention | Access | Storage Type | Cost/TB/Month | |------|-----------|--------|--------------|---------------| | Hot | 0-30 days | Instant | SSD | $20 | | Warm | 31-90 days | <1 min | HDD | $10 | | Cold | 91-365 days | <1 hour | Archive | $2 |
Cost Estimate: - Hot (35 TB): 35 × $20 = $700 - Warm (70 TB): 70 × $10 = $700 - Total: $1,400/month for call recordings
5.2 Transcript Storage¶
Transcript Storage Calculation:
Text transcripts (AI-generated):
├─ Daily transcripts: 10,000 calls
├─ Avg transcript size: 20 KB (8 min call, ~2,400 words)
├─ Daily storage: 10,000 × 20 KB = 200 MB/day
├─ Monthly storage: 200 MB × 22 = 4.4 GB/month
├─ Retention: 90 days
└─ Total required: 4.4 GB × 3 = 13.2 GB
Cost: Negligible (~$0.26/month for 13 GB)
5.3 Historical Reporting Data¶
Analyzer Historical Data:
Webex Analyzer data retention:
├─ Real-time data: 24 hours (in-memory)
├─ Historical reports: 13 months (Cisco-managed cloud)
├─ Custom exports: Store locally if >13 months needed
├─ Estimated data size: 500 GB/year
└─ Storage: Included in Webex license (Cisco-managed)
Custom Data Warehouse (if needed): - Export daily: 100 MB/day compressed - Annual storage: 36 GB/year - 5-year retention: 180 GB - Cost: ~$3.60/month (cold storage)
5.4 System Logs and Audit Logs¶
Log Storage:
CUBE, Flow Designer, Integration logs:
├─ CUBE logs: 50 MB/day
├─ Flow Designer logs: 100 MB/day
├─ Integration logs: 200 MB/day
├─ Webex CC platform logs: Cisco-managed (included)
├─ Total custom logs: 350 MB/day = 7.7 GB/month
├─ Retention (audit): 365 days
└─ Total required: 7.7 GB × 12 = 92.4 GB
Cost: ~$1.85/month (cold storage)
5.5 Backup and Disaster Recovery Storage¶
Configuration Backups:
Daily backups of:
├─ Flow Designer flows: 50 MB
├─ Queue/routing configs: 10 MB
├─ Agent profiles: 5 MB
├─ Total per backup: 65 MB
├─ Daily backups × 90 days: 65 MB × 90 = 5.85 GB
└─ Geo-redundant storage: 5.85 GB × 2 = 11.7 GB
Cost: ~$0.23/month
5.6 Storage Capacity Summary¶
Total Storage Requirements (1,000-agent deployment):
| Storage Type | Size | Retention | Monthly Cost | Annual Cost |
|---|---|---|---|---|
| Call Recordings | 105 TB | 90 days | $1,400 | $16,800 |
| Transcripts | 13 GB | 90 days | $0.26 | $3 |
| Historical Data | 36 GB/year | 5 years | $3.60 | $43 |
| System Logs | 92 GB | 365 days | $1.85 | $22 |
| Backups | 12 GB | 90 days | $0.23 | $3 |
| TOTAL | ~105 TB | Mixed | $1,406 | $16,871 |
📍 Also see: dr-and-resiliency.md for backup and DR storage strategies
6. Integration API Capacity¶
6.1 CRM Integration (Salesforce)¶
API Call Volume:
Screen pop and activity logging:
├─ Concurrent calls: 700
├─ Screen pop (inbound): 1 API call/call = 700 calls/hour
├─ Activity logging: 2 API calls/call = 1,400 calls/hour
├─ Total API calls/hour: 2,100 calls/hour
├─ API calls/minute: 35/min
└─ Salesforce limit: 100,000 API calls/day (Enterprise)
Status: ✅ Well within Salesforce limits
6.2 Workforce Management (WFM) Integration¶
Data Sync Volume:
Real-time adherence sync:
├─ Agent state updates: Every 30 seconds
├─ Agents monitored: 1,000
├─ Updates per hour: 1,000 × (60/0.5) = 120,000 updates/hour
├─ API calls (batched): 120,000 / 100 = 1,200 API calls/hour
└─ Calabrio API limit: No documented limit (cloud-native)
7. Capacity Monitoring and Alerting¶
7.1 Capacity Monitoring Thresholds¶
Set alerts at these thresholds:
| Component | Warning (%) | Critical (%) | Action |
|---|---|---|---|
| CUBE sessions | 70% | 85% | Add CUBE capacity |
| Network bandwidth | 70% | 85% | Upgrade circuits |
| Storage (call recordings) | 75% | 90% | Add storage tier |
| AI API quotas | 70% | 85% | Request quota increase |
| Database connections | 70% | 85% | Scale database |
7.2 Capacity Planning Review Cadence¶
Regular reviews: - Weekly: During initial migration (first 3 months) - Monthly: First year post-migration - Quarterly: Ongoing steady-state
8. Growth Planning and Scaling¶
8.1 3-Year Capacity Projection¶
Assumed growth: 15% annually
| Year | Agents | CUBE Sessions | Bandwidth | Storage (Monthly) | AI Costs (Monthly) |
|---|---|---|---|---|---|
| Y1 | 1,000 | 6,084 | 682 Mbps | 105 TB | $281,840 |
| Y2 | 1,150 (+15%) | 6,997 | 784 Mbps | 121 TB | $324,116 |
| Y3 | 1,323 (+15%) | 8,046 | 902 Mbps | 139 TB | $372,733 |
Key Scaling Decisions: - Year 2: Add 3rd CUBE (reach 10,500 total sessions) - Year 2: Upgrade internet to 750 Mbps per circuit - Year 3: Review storage tier strategy (move to cheaper cold storage)
9. Capacity Planning Tools¶
9.1 Available Calculators¶
Use these tools for capacity planning:
- CUBE Session Calculator → See cube-and-sbc-design.md Section 4.6
- Network Bandwidth Calculator → See network-architecture.md Section 11.3
- AI Cost Estimator → Use Google Cloud Pricing Calculator
- Storage Cost Calculator → AWS S3 Pricing Calculator or equivalent
9.2 Capacity Planning Spreadsheet¶
Request from architecture team: - Excel workbook with all formulas - Input: # of agents, growth %, usage patterns - Output: All capacity requirements and costs
10. Summary and Recommendations¶
10.1 Critical Capacity Actions¶
Must complete before go-live:
- ✅ CUBE Capacity: Verified in cube-and-sbc-design.md (2× ASR 1002-HX)
- ✅ Network Bandwidth: Verified in network-architecture.md (dual 500 Mbps)
- ✅ Agent Licensing: Verified in assumptions-and-dependencies.md (1,050 licenses)
- 🔴 Google Speech-to-Text Quota: Request increase to 250-300 concurrent streams
- 🟡 Storage: Provision 120 TB (20% buffer) for call recordings
- 🟡 AI Cost Budget: Allocate $280-300K/month for AI services
10.2 Capacity Planning Success Factors¶
Keys to successful capacity planning: - ✅ Use actual traffic data (not just estimates) - ✅ Include 20-30% buffer for peaks and growth - ✅ Monitor capacity utilization from Day 1 - ✅ Plan for 3-year growth (not just initial deployment) - ✅ Set up automated alerts at 70% thresholds - ✅ Review capacity quarterly
11. Document Cross-References¶
Detailed Capacity Planning (Existing Docs)¶
| Topic | Document | Section |
|---|---|---|
| CUBE Sessions | cube-and-sbc-design.md | Section 4 (380 lines) |
| Network Bandwidth | network-architecture.md | Section 11 |
| Agent Licensing | assumptions-and-dependencies.md | Section 4.2 |
| Agent Capacity | design-principles.md | Section 4.1 |
| DR Storage | dr-and-resiliency.md | Section 8 |
New Content (This Document)¶
| Topic | Section |
|---|---|
| IVR Capacity | Section 3 |
| AI/Automation Capacity | Section 4 (most comprehensive) |
| Storage Capacity | Section 5 |
| Integration APIs | Section 6 |